The Cascadia fault stretches from the northern end of Vancouver Island, Canada to the southern end of Oregon. Major subduction events (M 9+) occur along the fault at an average of every 200 to 500 years, with the last one occurring in the winter of 1700 AD (316 years ago) (Witter, et al., 2003). There are two major results to a M9+ subduction earthquake off of the Oregon cost. First, a tsunami will hit the coast 15 to 30 minutes after the earthquake, flooding most coastal communities. Second, the entire coastal range will subside, which means that many areas will remain underwater as a new coastline is defined (Cascadia Region Earthquake Workgroup, 2013).
Similar events in Japan, Chile, and Indonesia have resulted in tens to hundreds of thousands of deaths (Parwanto, et al., 2013). Because of the hundreds of year-long gaps between major Cascadia events, the coastal communities have been built with little preparation for such an event. Multiple plans and emergency preparedness programs are being implemented (Cascadia Rising, Tsunami Ready, and state regulations for new construction).
Part of the preparation process is to locate areas where emergency shelters and supplies could be best distributed. Such locations should 1) be near the affected areas; 2) be near areas of higher population; 3) be in a relatively flat location; and 4) have access to the interior of the state. Locating areas to place disaster recovery centers can be accomplished through the use of GIS multi criteria evaluation (MCE). This will layer each criteria of concern and return the best-fit locations.
This analysis looks at the broad-scale impact that a Cascadia subduction tsunami might cause to the state as a whole; including population displacement and damaged highways. It then takes a closer look at Coos County, Oregon to determine where local recovery zones might be best located. By using MCE, this same process can be applied to other coastal counties.
Methods
The state-wide analysis utilized the DOGAMI local tsunami zone, the 2010 US census data, the 2013 city limits, and the 2013 ODOT highway network. Population displacement was determined by using the centroids of each census block. Highway damage was determined by calculating the difference between the length of the highway as a whole and the length of damaged highway.
The MCE for Coos County was created from four raster layers: 1) population density; 2) distance from an interior-leading highway; 3) slope; and 4) distance from the tsunami inundation zone. All layers were matched to the cell size and extent of the Coos County 50m DEM. The population density layer was created using a point kernel density created from the census block centroids (using Silverman’s Rule to determine the search radius). Slope was rasterized by degree. In Coos County the single undamaged highway to the interior only reaches the community of Coquille before becoming damaged by the tsunami. Distances from the tsunami zone and the highway were created using Euclidian distance.
Model for Computing Population Density:
Model for Locating Undamaged Highways:
Model for Computing Distance from Highway:
Model for Computing Distance from Tsunami:
Model for Computing Slope:
The next phase in analysis normalized all four raster layers and reclassed them to values from 1 to 10, with 10 being the best choice and 1 being the worst (Table 1). The population density was reclassed to use quantile classes. The slope was reclassed to emphasize the range of acceptable slopes (Hancock et al., 2007), the rest were divided evenly between the remaining classes. Both distance layers were reclassed based on walking distances. The tsunami distance layer also adds a class (with a value of 1) to anything within 500 feet of the inundation zone in order to avoid immediate flood damages or secondary tsunami. (The models for reclassification are not shown due to an unfortunate software error) Once the layers were reclassed, the values for each cell were added together with different weights for each layer (Table 2). Slope was given the highest weight because placing an emergency center on a steep slope would not be feasible, no matter how well the other criteria matched. The tsunami distance and population density layers were given medium weights. The distance to the highway layer was given an almost non-existent weight because of its irrelevance to most of the county.
| NORMALIZATION | Population Density | Distance to Highway (ft) | Slope | Distance from Tsunami (ft) |
| Perfect Choice: 10 | 2,550 | 0-2,000 | 0-0.5⁰ | 2,000 |
| Good Choice: 9 | 500 | 2,000 | 1⁰ | 6,000 |
| Okay Choice: 8 | 230 | 6,000 | 2⁰ | 10,000 |
| Poor Choice: 7 | 130 | 10,000 | 4⁰ | 20,000 |
| Terrible, but workable: 6 | 80 | 20,000 | 5⁰ | 30,000 |
| Unworkable: 5 | 50 | 30,000 | 7⁰ | 71,000 |
| 4 | 30 | 104,000 | 17⁰ | 112,000 |
| 3 | 20 | 142,000 | 27⁰ | 153,000 |
| 2 | 0-10 | 179,000 | 37⁰ | 240,000 |
| So bad it is Ridiculous: 1 | 0 | 220,000 | 50⁰ | 0-500 |
| Normalization Method | Quantile | Travel Distance | ForestService* | Travel Distance |
Table 1: Breakpoints for the normalization of the MCE raster layers from 1 to 10.
* (Hancock et al., 2007)
| Input | Weight |
| Population density | 25% |
| Distance to Highway | 5% |
| Slope | 40% |
| Distance to Tsunami | 30% |
Table 2: Weights for each raster layer for the MCE.
Results
The state-wide effects of a local tsunami vary greatly between the seven counties of Oregon (Table 3). A total of about 56,300 people could be displaced when their homes are inundated, requiring emergency housing. 256 miles of highway along the coast would be damaged. This is especially concerning for counties that rely on their costal route to connect with the rest of the state.
| Population | Housing | Highway (miles) | |||||||
| Total | Displaced | % | Total | Damaged | % | Total | Damaged | % | |
| Clatsop | 37,039 | 18,963 | 51.2% | 21,546 | 12,750 | 59.2% | 176 | 48 | 27.2% |
| Coos | 63,043 | 9,787 | 15.5% | 30,593 | 6,215 | 20.3% | 172 | 57 | 33.1% |
| Curry | 22,364 | 6,375 | 28.5% | 12,613 | 3,876 | 30.7% | 112 | 30 | 27.2% |
| Douglas | 107,667 | 1,331 | 1.2% | 48,915 | 1,123 | 2.3% | 469 | 17 | 3.7% |
| Lane | 351,715 | 1,841 | 0.5% | 156,112 | 1,129 | 0.7% | 552 | 18 | 3.2% |
| Lincoln | 46,034 | 6,219 | 13.5% | 30,610 | 9,800 | 32.0% | 182 | 50 | 27.4% |
| Tillamook | 25,250 | 11,799 | 46.7% | 18,359 | 7,457 | 40.6% | 148 | 36 | 24.2% |
| Total | 653,112 | 56,315 | 8.6% | 318,748 | 42,350 | 13.3% | 1,811 | 256 | 14.1% |
Table 3: Summary of per-county effects of a local tsunami.
Coos County
Coos County has five coastal cities that would be affected by a local tsunami, as well as a sizable population living outside of the city borders. There is a single highway leading to the interior: from Coquille, in Coos County to Roseburg, in Douglas County. The terrain of Coos County is largely mountainous, with the majority of flat areas located near the coast. Unfortunately, most of these areas are also likely to be inundated by a local tsunami and would not be appropriate for an emergency center.
| Land Area Submerged mi2 | Population | Housing | |||||||
| Total | Submerged | % | Total | Displaced | % | Total | Damaged | % | |
| Bandon | 14,093 | 10,443.0 | 74.1% | 3,674 | 2,160 | 58.8% | 2,182 | 1,375 | 63.0% |
| Coos Bay | 51,568 | 8,616.8 | 16.7% | 17,967 | 4,108 | 22.9% | 8,489 | 2,064 | 24.3% |
| Coquille | 14,248 | 703.7 | 4.9% | 4,467 | 205 | 4.6% | 2,079 | 90 | 4.3% |
| Lakeside | 10,466 | 800.0 | 7.6% | 2,020 | 76 | 3.8% | 1,162 | 57 | 4.9% |
| North Bend | 20,359 | 7,810.5 | 38.4% | 11,737 | 1,486 | 12.7% | 5,324 | 697 | 13.1% |
| Total | 110,734 | 28,373.9 | 25.6% | 39,865 | 8,035 | 20.2% | 19,236 | 4,283 | 22.3% |
Table 4: Summary of Damages and Displacement in Coos County, per city.
The placement of emergency mass care facilities in Coos County is based on two types of data. First, proximity to the affected populations (based on distance to the tsunami zone and population density), and areas that are large and flat enough to accommodate temporary FEMA shelters, resident vehicles, and so forth. For this analysis, I split Coos County into four major zones, each corresponding to the major city within that zone. For each zone, I identified the best fit areas to place emergency facilities. Depending on the actual situation, these best fit areas might not actually be available, for example in the case of a more extreme tsunami event than anticipated, or the concurrence of other disasters such as earthquake debris, mudslides, or landslides in the areas.
Conclusion
The largest concern that presented itself in this analysis was the lack of highway leading to the hardest struck regions along the coast. Coos County has a single highway leading from the coast to the rest of the state, Curry County to the south doesn’t have any. Supplies and emergency care will likely need to be transported by ship or helicopter rather than by road—leading to the possibility of delayed response times and insufficient supplies.
The second particular that I want to emphasize is the different recovery strategies needed for larger cities such as North Bend/Coos Bay where only a small percentage of the city will be affected and coastal towns such as Bandon where almost 75% of the town might be destroyed. The larger towns will retain some internal infrastructure to assist with the damage and emergency care, cities like Bandon and Gearhart (which might be 100% underwater) will have to rely more heavily on external emergency services.
References
Cascadia Region Earthquake Workgroup. (2013). Cascadia Subduction Zone Earthquakes: A magnitude 9.0 earthquake scenario. p. 7. http://www.crew.org/sites/default/files/cascadia_subduction_scenario_2013.pdf
Hancock, J.; Vander Hoek, K. K. J.; Bradshaw, S., Coffman, J. D., and Engelmann, J. (2007). Equestrian Design Guidebook for Trails, Trailheads, and Campgrounds. U.S. Department of Agriculture, Forest Service, Missoula Technology and Development Center. http://www.fs.fed.us/t-d/pubs/pdfpubs/pdf07232816/pdf07232816dpi72pt08.pdf
Parwanto, Novia Budi, & Oyama, Tatsuo. (2013). A statistical analysis and comparison of historical earthquake and tsunami disasters in Japan and Indonesia. International Journal of Disaster Risk Reduction, 7, 122-141.
Witter, R. C., Kelsey, H. M., & Hemphill-Haley, E. (2003). Great Cascadia earthquakes and tsunamis of the past 6700 years, Coquille River estuary, southern coastal Oregon. Geological Society of America Bulletin, 115(10), 1289-1306.